Development of a CT image-based virtual atelectasis simulation model and noninvasive lung nodule localization system

Introduction

Atelectasis is an essential consideration in lung nodule removal. During the surgical procedure, gauging the location of the nodule using computed tomography (CT) is challenging (1,2). Since surgeons rely on limited two-dimensional (2D) information to estimate the location of nodules, there is a need to develop precise and accessible methods to estimate nodule locations, particularly within the constraints of atelectasis.

Implementing lung cancer screening CTs for high-risk groups has resulted in an increased detection rate of early-stage lung cancers, which are often characterized by subsolid nodules or small nodules measuring less than 3 cm (3). While percutaneous needle biopsy (PCNB) and transbronchial lung biopsy (TBLB) are methods for presurgical histological confirmation of lung nodules, these procedures present potential risk of complications, and a possibility of false negatives (4,5). This is particularly notable for small nodules located in the central area of the lungs without an open bronchus sign.

Consequently, the need for surgical biopsy using video-assisted thoracoscopic surgery (VATS) has become increasingly apparent (6). However, small nodules may shift from their locations on presurgical CT scans due to intraoperative atelectasis. Additionally, differentiating subsolid nodules from normal lung tissue by touch can be difficult, making their localization challenging. Currently, presurgical localization techniques, such as CT guided hook-wire localization, radioactivity, fluorescence, and dyes, are invasive and may pose additional risks of complications (7-10).

While CT has been used to create three-dimensional (3D) models to identify vascular and bronchial structures (11), no model simulation of atelectasis has been developed thus far. We aimed to develop a CT-based virtual atelectasis lung model (VALM) and compare it to VATS imaging to accurately reproduce atelectasis. We present this article in accordance with the TRIPOD reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-903/rc).

Methods

Clinical patient enrollment, screening, and clinical data acquisition

This study retrospectively reviewed 550 VATS procedures performed from June 28, 2021, to January 22, 2024, at Korea University Ansan Hospital. A total of 271 patients were excluded from the study for reasons such as under 19 years old, refusal to undergo surgery, transfer to other hospitals, low CT image quality, and mismatched image and video data. Out of the 279 patients who underwent surgery for lung nodules, 20 patients with either subsolid nodules or small nodules less than 3 cm in size were included in the study because they required localization at the discretion of the principal investigator (Figure 1). For nodules deeper than 2 cm from the lung surface or pure ground-glass nodules (GGNs), a hook-wire was used to mark the location of the nodule. In cases where other findings were easily identifiable, surgery was performed using conventional CT images alone.

Figure 1 Flowchart of the patient enrollment process of the study cohort.

The clinical data of patients were also reviewed through medical records. This study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). This study was approved by the Korea University College of Medicine Institutional Review Board (IRB No. 2023AS0299) and individual consent for this retrospective analysis was waived due to the retrospective nature.

3D reconstruction of inflated lungs

A CT scan was performed approximately one week before VATS. Presurgical contrast-enhanced chest CT scans were conducted using one of the multidetector row CT scanners; IQon (Philips Healthcare, Massachusetts, United States), Ingenuity Core 128 (Philips Healthcare, United States). The CT scanning employed specific parameters, including detector collimation, 0.625–40 mm; beam pitch, 0.609; reconstruction thickness, 3 mm; reconstruction interval, 3 mm; tube voltage, 100–120 kVp; tube current, 30–150 mAs; and the reconstruction kernel, standard soft-tissue reconstruction algorithm. A total of 100–120 mL of non-ionic low osmotic iodine contrast medium was administered intravenously at a rate of 2–3 mL/sec using a power injector (CT motion; ulrich medical, Germany) in all patients.

Conversion of 2D CT images into 3D models was a prerequisite for simulating atelectasis, as illustrated in Figure 2A. We used interactive lobe segmentation (ILS) within the chest imaging platform (CIP) of 3D Slicer (12,13), an open-source software module designed for medical image analysis. To distinguish each lung lobe, three to five points were placed on each CT image, separating left upper lobe (LUL), left lower lobe (LLL), right upper lobe (RUL), right middle lobe (RML), and right lower lobe (RLL). The model incorporated the lungs, thoracic cavity, sternum, ribs, and nodules, with separation between the left and right sides determined by the mediastinum. The nodule of interest was segmented and transformed into separate 3D objects. 3D Slicer transformed the 3D model into the Wavefront .obj format.

Figure 2 Virtual atelectasis simulation process. (A) CT images were converted into a 3D model with each lobe identified using 3D Slicer. Based on the surgical posture of the patient and the location of the nodule, the 3D lung model was rotated and vertically flipped. Employing the drape function, the sternum shape was captured. (B) Mesh points descend vertically until the concave sternum is reached with approximately 50% of the lung volume. (C) An image showcasing the lung shape is extracted from the surgery video. A virtual camera that is placed in the same location captures a simulated image that mirrors the surgery video. The similarity of the shape is evaluated, with a primary focus on the outline of the lung. Texture is applied to the 3D atelectasis model (Video 1), enabling real-time presentation of atelectasis images in the operating room via the Unity3D and Intel RealSense T265. DICOM, digital imaging and communications in medicine; SLAM, simultaneous localization and mapping; VATS, video-assisted thoracoscopic surgery; VALM, virtual atelectasis lung model; MATLAB, matrix laboratory; CT, computed tomography.

Execution of atelectasis simulation according to gravitational direction

The model was rotated to the same angle as the surgical position to facilitate the simulation (Figure 2A). To manipulate multiple mesh points within the constructed 3D model, Rhinoceros3D, an optimized comptuer-aided design (CAD) program (14), and Grasshopper (15), a visual programming language integrated into the Rhinoceros3D software, were used to handle extensive real-time calculations. Kangaroo Physics (16) was used as the physics engine for interactive modeling simulation within Grasshopper to shape the lung according to gravity (Figure S1).

To ensure swift simulation implementation, each mesh object was streamlined into objects containing 5,000 vectors. These 5,000 points descended (negative z-axis in Rhinoceros3D) at the default gravitational acceleration speed in Kangaroo (Figure 2B).

The morphology of the lung is influenced by their position within the thorax. Subsequently, a specific construct was developed to interrupt the descent simulation, tailored to the unique shape of each lung, and primarily dictated along the edge of the mediastinal configuration. To achieve this, we created a bowl designed to cradle the lungs in the surgical position. The 3D lung model underwent dynamic flipping, utilizing Rhinoceros3D’s “Drape” function to extract the cloth-like form draped over it (Figure 2A). The shape of the cloth effectively mimicked the mediastinal contours. By simulating the gravitational direction, each vector point within the lung trajectory was meticulously positioned along the movement path and halted upon contact (Figure 2B). Independent simulations were systematically conducted for the upper, middle, and lower lobes and the lung nodules.

Each lung section contained approximately 5,000 vector points and 10,000 faces; all vector points were moved in the direction of gravity by a specified length per frame until they reached approximately 49% of the total lung volume, which is the average observed during this study (Table S1). The shape of the simulated atelectasis flattened as the process advanced (Video 2). The contraction rate is readjust based on the atelectatic lung image generated according to the patient’s respiratory status after anesthesia at the operation room.

Statistical analysis

Comparative analysis of simulation results with VATS imaging similarity coefficient

To quantify the similarity between the images captured by the VATS camera and those captured by the simulation camera in the Rhinoceros 3D program, we used three established statistical methods: the Dice Similarity Coefficient (DSC), the Jaccard Similarity Coefficient (JSC), and the Hausdorff distance (HD). These methods were employed to validate the accuracy and efficacy of the 3D lung atelectasis simulation model. Our model used DSC and JSC to measure the overlap between the lung surface identified in surgical video snapshots and their counterparts in the simulated 3D model (Figure 2C and Video 3). The DSC and JSC formula as follow:

DSC=2|A∩B|A+B[1]

JSC=|A∩B||A∪B|[2]

In the formula, set A represents the atelectasis surface areas in surgical video snapshots, while set B describes the atelectasis surface areas as predicted by the simulated 3D models. The calculation of DSC and JSC values, which range from zero (no overlap) to one (perfect overlap), serves to measure the extent to which the 3D model faithfully replicates the extent and position of lung atelectasis in surgical video snapshots. Consequently, higher values signify a greater degree of similarity. This assessment is achieved through a comparison of the overlap between the lung atelectasis region identified in the video snapshot and the region predicted by the 3D model. The high DSC values underscore the model’s precision in capturing the exact dimensions and positions of the atelectasis regions. Similarly, the JSC evaluates the similarity and variance between the two datasets by calculating the area of their intersection relative to the size of their union. The JSC is especially notable, as it provides an analysis that complements the findings obtained from the shape similarity between the lungs in video snapshot and the 3D simulations. Whereas the DSC is predominantly focused on accurately replicating the atelectasis areas in terms of extent and position, the JSC offers a wider perspective, emphasizing the shape and diversity of these zones.

Spatial accuracy

HD is a metric used to measure the spatial accuracy between two sets of points by determining the maximum distance between the points on the margin of the nodule in one set, such as surgical video images, and their corresponding points in the other set, such as simulated images. Unlike similarity coefficients like DSC and JSC, which quantify overlap and shape similarity, HD specifically evaluates how far apart corresponding features, such as the location of nodules, are in the two images, providing insight into the geometric accuracy of the simulation. A smaller HD indicates a closer geometric match and higher spatial precision. Given two sets A and B, within a Euclidean space, consisting of points, the H (A,B) from A to B defined as follows:

H(A,B)=max{supa∈Ainfb∈Bd(a,b),supb∈Binfa∈Ad(a,b)}[3]

d (a,b) represents the Euclidean distance between the points a and b. The sup (supremum) denotes the greatest possible distance within a set, indicating the maximum of all distances considered. Conversely, inf (infimum) refers to the smallest distance from a point in one set to any point in the other set, capturing the minimum of these distances. That is, HD evaluates the maximum distance of the closest point and the minimum distance of the farthest point to quantify the degree of similarity or inconsistency between the two sets.

Correlation analysis

To quantitatively measure how well the changes in the metrics (DSC, JSC, and HD) of our simulation model match the changes observed in the corresponding metrics of video snapshots, we also evaluated the Pearson correlation coefficient (PCC). The PCC serves as a statistical measure that quantifies the strength and direction of a linear relationship between two continuous variables. Its value ranges from −1 to +1, where +1 denotes a perfect positive linear relationship, 0 signifies no linear correlation, and −1 indicates a perfect negative linear relationship. High PCCs show that simulations accurately maintain the observed quantitative relationships between different measurements in clinical data. All statistical analyses comparing simulation results with VATS imaging were conducted using MATLAB R2021b (MathWorks, Natick, MA, USA).

Correlation analysis of simulation data with lung functions parameters

The study explored the relationship between the extent of lung atelectasis and baseline pulmonary function parameters in patients. To achieve this, the forced expiratory volume in one second (FEV1)/forced vital capacity (FVC) ratios and FVC values associated with obstructive or restrictive lung diseases were collected. These were then correlated with volumetric measures of lung shrinkage obtained from simulations. The PCC was calculated using IBM SPSS Statistics for Windows, Version 26.0 (IBM Corp., Armonk, NY, USA).

Results

Demographics

In the analysis, 20 patients were included, with a sex ratio of 9 males to 11 females and an average age of 61 years [standard deviation (SD), ±9.96]. An average nodule size of 14.8 mm. And the morphology of the nodules was characterized as follows: eight ground-grass, six part-solid, and six solid nodules. The anatomical distribution of lung nodules was as follows: six cases in the RUL, one case in the RML, five cases in the RLL, four cases in the LUL, and four cases the LLL. Among these, preoperative hook-wire localization was performed in six cases (Table 1).

Simulation result

The average volume of the entire lung was 2,336 cm³ (±588). During VALM, the inferior surface of lung (central area of the lung) maintained their initial shapes, while superior surface of lung (peripheral area of the lung) became contracted or distorted based on the shape of the previously created “Bowl” (resemble mediastinal contours) (Figure 2A and Table S2). The final volume of the lung atelectasis measured 1,181 cm³ (±369), and the average shrinkage rate was 48.6% (±12.9%) (Figure 3 and Table S1). The results of pre-operative pulmonary function tests in patients showed no significant correlation with lung atelectasis degree, as demonstrated in Table S1.

Figure 3 Simulation results. (A) Patient No. 1 (RUL)—axial and coronal view with nodule location (green lines), 3D models of inflated and deflated lung (gray: right superior lobe, blue: right middle lobe, green: right inferior lobe, yellow: nodule). (B) Patient No. 2 (LUL)—axial and coronal view with nodule location (green lines), 3D models of inflated and deflated lung (grey: left superior lobe, green: left inferior lobe, yellow: nodule). (C) Patient No. 3 (RUL)—axial and coronal view with nodule location (green lines), 3D models of inflated and deflated lung (bronze: right superior lobe, blue: right middle lobe, gray: right inferior lobe, yellow: nodule). (D) Patient No. 7 (LLL)—axial and coronal view with nodule location (green lines), 3D models of inflated and deflated lung (bronze: left superior lobe, gray: left inferior lobe, yellow: nodule). (E) Patient No. 19 (RML)—axial and coronal view with nodule location (green line), 3D models of inflated and deflated lung (bronze: right superior lobe, gray: right middle lobe, green: right inferior lobe, yellow: nodule). VALM, virtual atelectasis lung model; RUL, right upper lobe; LUL, left upper lobe; LLL, left lower lobe; RML, right middle lobe.

In this study, a comparative analysis was conducted between instantaneously captured images of atelectasis simulation results and snapshot pictures from surgical videos across 20 cases (Video 3). On average, 15 images were compared for each case. When assessing the conformity of the lung surface area, the average DSC value was 90.27, and the average JSC value was 88.25, indicating a high degree of agreement. In addition, the congruency between the nodule locations within the simulation and those anticipated in the surgical video was calculated at an average HD of 6.39 mm (Table 2).

Discussion

This study developed VALM to aid in noninvasive lung nodule localization, aiming to improve presurgical planning precision. Among 279 patients undergoing VATS surgery for lung nodules, 20 with small nodules were studied, with CT images converted into 3D models for simulation. The simulation demonstrated close resemblance to surgical videos, indicating potential for enhancing nodule location accuracy and improving surgical planning efficiency (Video 4).

The incidence of lung cancer is gradually increasing globally. As of 2019, the prevalence of lung cancer in South Korea has continued to increase, with the number of patients increasing from 24,757 to 30,241 in recent years (17). The National Cancer Screening Project conducts biennial low-dose chest CT scans for smokers aged 54–74 years with a history of smoking more than 30 packs per year, thereby enhancing the detection rate of lung nodules and early-stage lung cancer. With the increasing discovery of smaller subsolid lung nodules, implementing VATS for malignancy confirmation is increasing (18,19).

Before performing VATS, the surgeon must reconstruct and visualize the location of the nodule in 3D space. Recently, software like Synapse Vincent has enabled quick execution of 3D transformation (20,21). The task includes localization of the nodule using this information and matching it with the 2D image observed through the thoracoscope. Nonetheless, because CT scans are performed in an inspiratory state, tracking the location of the nodule becomes challenging during surgery when air is removed from the lung (22). Additionally, intraoperative postural changes lead to alterations in the direction of lung descent owing to atelectasis.

The shape of the lung continues to change during surgery. The lungs show a large difference between the inspiratory and expiratory states. Anesthesia administered prior to general surgical procedures, has been reported to induce lung collapse of up to 15–20% (23). However, there are no reports on the extent of lung atelectasis during thoracoscopic surgery. In our study, we directly calculate the degree of lung atelectasis using preoperative CT and intraoperative video data. Although there was variability among patients, we observed an average volume shrinkage of 49% during thoracoscopic surgery.

Owing to the characteristics of the lungs, when air is blown out, the shape flattens in the direction of gravity, depending on the patient’s surgical posture (24). For resection, the entire lung on the side of the nodule must be in a minimally aerated state of atelectasis. Therefore, the location of the nodule is predicted to sink at a certain rate, depending on the degree of lung descent. Because the preoperative CT image is captured in the inspiratory state, the shape of the lung and the location of the nodule recognized by the operator based on the preoperative CT image are significantly different from the image seen during the actual surgery. Previous research has produced models of lung atelectasis, but none that simulate atelectasis from a surgeon’s viewpoint. While there are studies using a canine model for atelectasis, their relevance to human applications remains restricted (25-28).

Our research is the first to simultaneously simulate atelectasis and nodule movement in the human lungs. Based on the 3D lung model, a force in the direction of gravity was applied according to the patient’s surgical posture to create an atelectasis model, and the location of the nodule was moved according to the shape of the atelectasis. The system developed in this study shows the shape of the lungs closest to what the operator can observe through a thoracoscope during surgery with average DSC and JSC values of 90.27 and 88.25. This allows us to visualize lung atelectasis and analysis quantitatively before the surgery and shows the reliable estimated location of the nodule as a previously unavailable noninvasive method.

Our study had limitations. First, the simulation of atelectasis lacks vascular and bronchial structures, potentially restricting the ability to establish landmarks when identifying nodules. Second, because it is difficult to accurately measure length using surgical images alone, the location of the nodule confirmed through the image may differ slightly from the actual location. Finally, the validity of the simulation was not confirmed during actual operations. In the future, it is expected that a research process will be needed to develop a simulation model by integrating vascular and bronchial structures and apply it to an actual operating room to verify the accuracy of the simulation and the accuracy of the nodule location. This will enable accurate replication of real-world anatomy and reliable prediction of nodule location, increasing clinical utility and ultimately improving surgical outcomes and patient safety.

Conclusions

We developed VALM and confirmed comparable results through VATS images. By predicting the location of nodules through a non-invasive method, it is expected to reduce the burden of additional invasive procedures on patients and aim to minimize lung resection. This system may aid in the patient’s rapid recovery and reduce the surgery time and likelihood of surgery-associated complications.

Acknowledgments

Funding: This study was supported by the National Research Foundation of Korea grant funded by the Korea government (Ministry of Science and ICT) (No. 2023R1A2C3005944, to I.H. and S.H.).

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-903/rc

Data Sharing Statement: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-903/dss

Peer Review File: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-903/prf

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-903/coif). I.H. and S.H. report funding from the National Research Foundation of Korea grant funded by the Korea government (Ministry of Science and ICT, No.2023R1A2C3005944). I.H., S.H.L., and J.H. report regarding this research, the domestic patent approval process is pending, the PCT application and the US application has been completed. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki (as revised in 2013) and approved by the Korea University College of Medicine Institutional Review Board [IRB approval number 2023AS0299, October 10, 2023 (initial)] and individual consent for this retrospective analysis was waived due to the retrospective nature.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Greenfield AL, Steiner RM, Liu JB, et al. Sonographic guidance for the localization of peripheral pulmonary nodules during thoracoscopy. AJR Am J Roentgenol 1997;168:1057-60. [Crossref] [PubMed]
2. Saito H, Minamiya Y, Matsuzaki I, et al. Indication for preoperative localization of small peripheral pulmonary nodules in thoracoscopic surgery. J Thorac Cardiovasc Surg 2002;124:1198-202. [Crossref] [PubMed]
3. Li C, Wang H, Jiang Y, et al. Advances in lung cancer screening and early detection. Cancer Biol Med 2022;19:591-608. [Crossref] [PubMed]
4. Lee SB, Kim MJ, Lee IJ. Assessment of diagnostic accuracy and complication rates of CT-guided percutaneous core-needle biopsy for lung lesion: difference between solid and sub-solid nodules based on propensity score matching analysis. Clin Radiol 2023;78:e620-6. [Crossref] [PubMed]
5. Philip B, Jain A, Wojtowicz M, et al. Current investigative modalities for detecting and staging lung cancers: a comprehensive summary. Indian J Thorac Cardiovasc Surg 2023;39:42-52. [Crossref] [PubMed]
6. Choi SM, Heo EY, Lee J, et al. Characteristics of benign solitary pulmonary nodules confirmed by diagnostic video-assisted thoracoscopic surgery. Clin Respir J 2016;10:181-8. [Crossref] [PubMed]
7. Ferguson MK, Vigneswaran WT. Diffusing capacity predicts morbidity after lung resection in patients without obstructive lung disease. Ann Thorac Surg 2008;85:1158-64; discussion 1164-5. [Crossref] [PubMed]
8. Chella A, Lucchi M, Ambrogi MC, et al. A pilot study of the role of TC-99 radionuclide in localization of pulmonary nodular lesions for thoracoscopic resection. Eur J Cardiothorac Surg 2000;18:17-21. [Crossref] [PubMed]
9. Anayama T, Qiu J, Chan H, et al. Localization of pulmonary nodules using navigation bronchoscope and a near-infrared fluorescence thoracoscope. Ann Thorac Surg 2015;99:224-30. [Crossref] [PubMed]
10. Sato M, Omasa M, Chen F, et al. Use of virtual assisted lung mapping (VAL-MAP), a bronchoscopic multispot dye-marking technique using virtual images, for precise navigation of thoracoscopic sublobar lung resection. J Thorac Cardiovasc Surg 2014;147:1813-9. [Crossref] [PubMed]
11. Xue L, Fan H, Shi W, et al. Preoperative 3-dimensional computed tomography lung simulation before video-assisted thoracoscopic anatomic segmentectomy for ground glass opacity in lung. J Thorac Dis 2018;10:6598-605. [Crossref] [PubMed]
12. 3D slicer image computing platform [Internet]. [cited 2024 Mar 15]. Available online: https://www.slicer.org/
13. Fedorov A, Beichel R, Kalpathy-Cramer J, et al. 3D Slicer as an image computing platform for the Quantitative Imaging Network. Magn Reson Imaging 2012;30:1323-41. [Crossref] [PubMed]
14. Associates RM, Rhinoceros 3D. [cited 2024 Mar 15]. Available online: https://www.rhino3d.com/
15. Davidson S. [cited 2024 Mar 15]. Available online: https://www.grasshopper3d.com/
16. Piker D. KANGAROO3D. [cited 2024 Mar 15]. Available online: http://kangaroo3d.com/
17. National Cancer Center. Annual report of cancer statistics in Korea 2019. Sejong: Ministry of Health and Welfare; 2021.
18. Lee YS, Kim JD, Park HO, et al. Video-Assisted Thoracic Surgery Core Needle Biopsy for Pulmonary Nodules in Patients with Impaired Lung Function: Is It Feasible and Safe? J Chest Surg 2023;56:1-5. [Crossref] [PubMed]
19. Ost DE, Gould MK. Decision making in patients with pulmonary nodules. Am J Respir Crit Care Med 2012;185:363-72. [Crossref] [PubMed]
20. Liu Y, Zhang S, Liu C, et al. Three-dimensional reconstruction facilitates thoracoscopic anatomical partial lobectomy by an inexperienced surgeon: a single-institution retrospective review. J Thorac Dis 2021;13:5986-95. [Crossref] [PubMed]
21. Ujiie H, Yamaguchi A, Gregor A, et al. Developing a virtual reality simulation system for preoperative planning of thoracoscopic thoracic surgery. J Thorac Dis 2021;13:778-83. [Crossref] [PubMed]
22. Wang G, Lin Y, Zheng L, et al. A new method for accurately localizing and resecting pulmonary nodules. J Thorac Dis 2020;12:4973-84. [Crossref] [PubMed]
23. Hedenstierna G, Edmark L. Mechanisms of atelectasis in the perioperative period. Best Pract Res Clin Anaesthesiol 2010;24:157-69. [Crossref] [PubMed]
24. Woodring JH, Reed JC. Types and mechanisms of pulmonary atelectasis. J Thorac Imaging 1996;11:92-108. [Crossref] [PubMed]
25. Wu S, Nakao M, Matsuda T. Continuous lung region segmentation from endoscopic images for intra-operative navigation. Comput Biol Med 2017;87:200-10. [Crossref] [PubMed]
26. Nakao M, Tokuno J, Chen-Yoshikawa T, et al. Surface deformation analysis of collapsed lungs using model-based shape matching. Int J Comput Assist Radiol Surg 2019;14:1763-74. [Crossref] [PubMed]
27. Lesage AC, Rajaram R, L, Tam A, et al. Preliminary evaluation of biomechanical modeling of lung deflation during minimally invasive surgery using pneumothorax computed tomography scans. Phys Med Biol 2020;65:225010. [Crossref] [PubMed]
28. Zhang R, Ferguson MK. Video-Assisted versus Open Lobectomy in Patients with Compromised Lung Function: A Literature Review and Meta-Analysis. PLoS One 2015;10:e0124512. [Crossref] [PubMed]